In a series of 1,332 pregnancies in women with diabetes, perinatal fetal mortality varied in a statistically significant degree, with maternal factors and pregnancy complications expressed respectively, by the White and PBSP classifications. Also, mortality declined steadily over the years 1946 to 1972. Fatal congenital malformation was the most important single cause of perinatal death in recent years. A controlled trial must take into account the year of admission as well as the White and PBSP classifications; with the prevailing low and decreasing mortality, even large centers may be unable to fulfill the requirements as to sample size within a few years. Therefore, criteria other than perinatal mortality may be needed to assess the value of changes in treatment.